
import numpy as np                                                         #importing numpy                             
import pandas as pd                                                        #importing pandas
import seaborn as sns                                                      #import seaborn
import matplotlib.pyplot as plt                                            #import matplotlib
from sklearn import metrics                                                #import metrics
from sklearn.linear_model import LogisticRegression                        #import Logistic Regression
from sklearn.model_selection import train_test_split                       #import train_test
from sklearn.metrics import classification_report, confusion_matrix        #importing classification_report, confusion_matrix
from sklearn.tree import DecisionTreeClassifier                            #importing DecisionTree
from sklearn import tree                                                   #importing tree
from sklearn.svm import SVC                                                #importing SVC
from sklearn.model_selection import GridSearchCV                           #importing GridSearch
from sklearn.cluster import KMeans                                         #importing KMeans
%matplotlib inline


df = pd.read_csv('vs.csv')                                  #load the csv file and make the data frame
df.head()                                                   #previewing


nulldata = pd.DataFrame(df.isnull().sum())           #checking for null values in the data
nulldata.columns = ['Null values']                   #Remaning the column
nulldata                                             #previewing the data


df.info()                                                    #To show the detailed summary

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 846 entries, 0 to 845
Data columns (total 19 columns):
 #   Column                       Non-Null Count  Dtype 
---  ------                       --------------  ----- 
 0   compactness                  846 non-null    int64 
 1   circularity                  846 non-null    int64 
 2   distance_circularity         846 non-null    int64 
 3   radius_ratio                 846 non-null    int64 
 4   pr.axis_aspect_ratio         846 non-null    int64 
 5   max.length_aspect_ratio      846 non-null    int64 
 6   scatter_ratio                846 non-null    int64 
 7   elongatedness                846 non-null    int64 
 8   pr.axis_rectangularity       846 non-null    int64 
 9   max.length_rectangularity    846 non-null    int64 
 10  scaled_variance              846 non-null    int64 
 11  scaled_variance.1            846 non-null    int64 
 12  scaled_radius_of_gyration    846 non-null    int64 
 13  scaled_radius_of_gyration.1  846 non-null    int64 
 14  skewness_about               846 non-null    int64 
 15  skewness_about.1             846 non-null    int64 
 16  skewness_about.2             846 non-null    int64 
 17  hollows_ratio                846 non-null    int64 
 18  class                        846 non-null    object
dtypes: int64(18), object(1)
memory usage: 125.7+ KB


df.describe()                                                #Analyze the distribution of the dataset


df.dtypes                                                    #It shows data types of columns

compactness                     int64
circularity                     int64
distance_circularity            int64
radius_ratio                    int64
pr.axis_aspect_ratio            int64
max.length_aspect_ratio         int64
scatter_ratio                   int64
elongatedness                   int64
pr.axis_rectangularity          int64
max.length_rectangularity       int64
scaled_variance                 int64
scaled_variance.1               int64
scaled_radius_of_gyration       int64
scaled_radius_of_gyration.1     int64
skewness_about                  int64
skewness_about.1                int64
skewness_about.2                int64
hollows_ratio                   int64
class                          object
dtype: object


df.shape                                                     #It shows the number of rows and columns

(846, 19)


df.columns                                                   #It shows the columns in a dataset

Index(['compactness', 'circularity', 'distance_circularity', 'radius_ratio',
       'pr.axis_aspect_ratio', 'max.length_aspect_ratio', 'scatter_ratio',
       'elongatedness', 'pr.axis_rectangularity', 'max.length_rectangularity',
       'scaled_variance', 'scaled_variance.1', 'scaled_radius_of_gyration',
       'scaled_radius_of_gyration.1', 'skewness_about', 'skewness_about.1',
       'skewness_about.2', 'hollows_ratio', 'class'],
      dtype='object')


axes = df.boxplot(figsize= (13,5))                           #boxplot
lab = axes.get_xticklabels()                             
axes.set_xticklabels(lab, rotation=90)                       #rotating the labels by 90 degrees 
plt.show()


sns.countplot(x = 'class', data = df)                        #countplot for variable class
plt.show()


df.hist(figsize=(15,15))                                     #hist plot
plt.show()


plt.figure(figsize=(20,5))                                                 
sns.countplot(df['skewness_about'],hue= df['class'])                        #countplot for skweness_about and class
plt.show()

C:\Users\YASASWINI DESU\anaconda3\lib\site-packages\seaborn\_decorators.py:36: FutureWarning: Pass the following variable as a keyword arg: x. From version 0.12, the only valid positional argument will be `data`, and passing other arguments without an explicit keyword will result in an error or misinterpretation.
  warnings.warn(


sns.histplot(data=df, x="skewness_about", hue="class",palette='Dark2')      #histplot for skweness_about and class
plt.show()


plt.figure(figsize=(15,7))
sns.heatmap(df.corr(), annot = True, cmap = 'PuBuGn')                      #heatmap of correlation
plt.show()


sns.pairplot(df,hue='class')                                           #pairplot

<seaborn.axisgrid.PairGrid at 0x23a629d9c40>


Y = df['class']                                                            #copying class column to Y
df.drop('class', axis = 1,inplace = True)                                  #dropping the column class


X_train, X_test, y_train, y_test = train_test_split(df,Y, test_size=0.3)   #splitting data into train and test


model = LogisticRegression(max_iter =900)                                  #Using LogisticRegression
model.fit(X_train,y_train)

C:\Users\YASASWINI DESU\anaconda3\lib\site-packages\sklearn\linear_model\_logistic.py:763: ConvergenceWarning: lbfgs failed to converge (status=1):
STOP: TOTAL NO. of ITERATIONS REACHED LIMIT.

Increase the number of iterations (max_iter) or scale the data as shown in:
    https://scikit-learn.org/stable/modules/preprocessing.html
Please also refer to the documentation for alternative solver options:
    https://scikit-learn.org/stable/modules/linear_model.html#logistic-regression
  n_iter_i = _check_optimize_result(

LogisticRegression(max_iter=900)


predictions = model.predict(X_test)                                        #Predicting
#predictions


cnf_matrix=confusion_matrix(y_test,predictions)                            #Confusion matrix for testdata and predicted data
cnf_matrix

array([[65,  0,  0,  2],
       [ 2, 35, 25,  3],
       [ 3, 11, 46,  0],
       [ 1,  0,  0, 61]], dtype=int64)


sns.heatmap(cnf_matrix, annot=True, fmt='g',cmap='gnuplot')               #heapmap for confusion matrix

<AxesSubplot:>


#Classification report measures the quality of predictions
print(classification_report(y_test,predictions))                           #printing the classification report

              precision    recall  f1-score   support

         bus       0.92      0.97      0.94        67
        opel       0.76      0.54      0.63        65
        saab       0.65      0.77      0.70        60
         van       0.92      0.98      0.95        62

    accuracy                           0.81       254
   macro avg       0.81      0.81      0.81       254
weighted avg       0.81      0.81      0.81       254


#accuracy, recall value, precision value
print("Accuracy Score is:",metrics.accuracy_score(y_test,predictions))                        
print('Recall value: ',metrics.recall_score(y_test, predictions, average='macro'))
print('Precision value: ',metrics.precision_score(y_test, predictions, average='macro'))

Accuracy Score is: 0.8149606299212598
Recall value:  0.814787106650371
Precision value:  0.8121230677874891


Logistic_Regression_Accuracy =  metrics.accuracy_score(y_test,predictions)  #accuracy of logistic regression


compare_logistic = pd.DataFrame({"Actual":y_test,"Predicted":predictions})  #actual vs predicted
compare_logistic


X_train, X_test, y_train, y_test = train_test_split(df,Y, test_size=0.30)  #training the model


dec_model = DecisionTreeClassifier(max_depth = 6,random_state = 0,criterion = "entropy")      #decision tree classifier
dec_model.fit(X_train,y_train)

DecisionTreeClassifier(criterion='entropy', max_depth=6, random_state=0)


Predictions=model.predict(X_test)                                   #predicting
#Predictions


print(confusion_matrix(y_test, Predictions))                         #confusion matrix

[[54  1  0  1]
 [ 0 41 20  1]
 [ 6 20 47  2]
 [ 1  0  0 60]]


print(classification_report(y_test, Predictions))                    #classification report

              precision    recall  f1-score   support

         bus       0.89      0.96      0.92        56
        opel       0.66      0.66      0.66        62
        saab       0.70      0.63      0.66        75
         van       0.94      0.98      0.96        61

    accuracy                           0.80       254
   macro avg       0.80      0.81      0.80       254
weighted avg       0.79      0.80      0.79       254


Decision_Tree_Accuracy = metrics.accuracy_score(y_test,Predictions)  #printing the accuarcy of decision tree classifier


compare_decision = pd.DataFrame({"Actual":y_test,"Predicted":Predictions})  #actual vs predicted
compare_decision


fig, axes = plt.subplots(nrows = 1,ncols = 1,figsize = (100,50), dpi = 400)   #decision tree
tree.plot_tree(dec_model,filled = True, rounded=True);


svc_model = SVC()                                                             #SVM Classifier
svc_model.fit(X_train,y_train)

SVC()


predict = svc_model.predict(X_test)                                           #predicting
#predict


print(confusion_matrix(y_test, predict))                                      #printing the confusion matrix

[[40  9  5  2]
 [18 23  7 14]
 [16 35  9 15]
 [17  0  0 44]]


print(classification_report(y_test, predict))                                #printing the classification report

              precision    recall  f1-score   support

         bus       0.44      0.71      0.54        56
        opel       0.34      0.37      0.36        62
        saab       0.43      0.12      0.19        75
         van       0.59      0.72      0.65        61

    accuracy                           0.46       254
   macro avg       0.45      0.48      0.43       254
weighted avg       0.45      0.46      0.42       254


SVC_Accuracy = metrics.accuracy_score(y_test,predict)                          #printing the accuarcy of decision tree classifier


compare_svc = pd.DataFrame({"Actual":y_test,"Predicted":predict})              #actual vs predicted
compare_svc


# Dictionary containing hyperparameters to be tuned and their values to be tested
grid_parameters = {'C': [0.001,0.1,1, 10, 100, 1000], 'gamma': [0.0001, 0.001, 0.01, 0.1, 1, 10, 100]}


grid = GridSearchCV(SVC(),grid_parameters,verbose=1)                          #gridsearchcv


grid.fit(X_train,y_train)

Fitting 5 folds for each of 42 candidates, totalling 210 fits

GridSearchCV(estimator=SVC(),
             param_grid={'C': [0.001, 0.1, 1, 10, 100, 1000],
                         'gamma': [0.0001, 0.001, 0.01, 0.1, 1, 10, 100]},
             verbose=1)


grid_prediction = grid.predict(X_test)                             #predicting


print(confusion_matrix(y_test,grid_prediction))                    #printing the confusion matrix

[[56  0  0  0]
 [ 1 42 17  2]
 [ 3 28 42  2]
 [ 0  1  1 59]]


print(classification_report(y_test,grid_prediction))               #printing the classification report

              precision    recall  f1-score   support

         bus       0.93      1.00      0.97        56
        opel       0.59      0.68      0.63        62
        saab       0.70      0.56      0.62        75
         van       0.94      0.97      0.95        61

    accuracy                           0.78       254
   macro avg       0.79      0.80      0.79       254
weighted avg       0.78      0.78      0.78       254


compare_svm = pd.DataFrame({"Actual":y_test,"Predicted":grid_prediction})        #actual vs predicted
compare_svm


SVM_Accuracy = metrics.accuracy_score(y_test, grid_prediction)           #svm accuracy after tuning


X_train, X_test, y_train, y_test = train_test_split(df,Y, test_size=0.30) #splitting into test and train


X = df.loc[:, ['compactness', 'circularity', 'distance_circularity', 'radius_ratio',   #all the columns
       'pr.axis_aspect_ratio', 'max.length_aspect_ratio', 'scatter_ratio',
       'elongatedness', 'pr.axis_rectangularity', 'max.length_rectangularity',
       'scaled_variance', 'scaled_variance.1', 'scaled_radius_of_gyration',
       'scaled_radius_of_gyration.1', 'skewness_about', 'skewness_about.1',
       'skewness_about.2', 'hollows_ratio']]


plt.figure(figsize=(15,7))
sns.set_style('darkgrid')
disortions = []                                                
for k in range(1, 10):                                         #We want k values from 1 to 10
    kmeans = KMeans(n_clusters = k)                            #initializing KMeans
    kmeans.fit(X)                                              #clustering
    disortions.append(kmeans.inertia_)                         #Inertia is the sum of squared error for each cluster
    
plt.scatter(range(1, 10), disortions ,c = 'g', s = 100)        #scattering for better reference
plt.plot(range(1, 10), disortions, c = 'r', linewidth = 3)     #plot an elbow model 
plt.title('Elbow Plot')                                        #giving title for the plot
plt.xlabel('Value of K')                                       #Adding the label for x-axis
plt.ylabel('Disortions')                                       #Adding the label for y-axis
plt.show()                                                     #previewing the plot

C:\Users\YASASWINI DESU\anaconda3\lib\site-packages\sklearn\cluster\_kmeans.py:881: UserWarning: KMeans is known to have a memory leak on Windows with MKL, when there are less chunks than available threads. You can avoid it by setting the environment variable OMP_NUM_THREADS=4.
  warnings.warn(


km = KMeans(n_clusters = 4, random_state=101)                   #Number of clsuters is 4
y = km.fit_predict(X)


cluster = km.labels_                                            #getting the values
#cluster = pd.DataFrame(data = cluster, columns = ['Cluster'])  #dataframe for clusters
#cluster.head()
#cluster


centers = km.cluster_centers_                                   #getting centers of final cluster
centers                                                         #previewing

array([[ 89.20612813,  42.77715877,  75.68523677, 158.2367688 ,
         62.91364903,   8.80779944, 153.50417827,  43.55431755,
         19.29805014, 143.79665738, 176.1810585 , 350.53760446,
        165.77158774,  73.61281337,   6.        ,  10.99721448,
        188.35097493, 194.8718663 ],
       [104.37234043,  53.48404255, 102.84042553, 200.57978723,
         61.66489362,   9.77659574, 219.62765957,  30.39361702,
         24.61702128, 167.71276596, 232.13829787, 715.46276596,
        216.25      ,  73.19148936,   7.37765957,  15.95212766,
        187.5212766 , 195.78723404],
       [ 98.06870229,  46.70229008,  93.21374046, 198.40458015,
         64.02290076,   8.59541985, 187.45801527,  35.06870229,
         21.90076336, 150.19847328, 207.55725191, 530.7480916 ,
        183.26717557,  68.29007634,   6.06870229,  13.38931298,
        193.10687023, 200.03816794],
       [ 87.8452381 ,  38.23214286,  63.875     , 133.43452381,
         57.30357143,   6.67857143, 130.25595238,  51.70238095,
         17.78571429, 133.20238095, 151.76190476, 251.70833333,
        140.61904762,  72.44047619,   6.30357143,  11.6547619 ,
        188.5       , 193.64880952]])


plt.figure(figsize=(15,7))
sns.set_style('white')                                                                            #setting style as white
plt.scatter(X.iloc[y == 0, 0], X.iloc[y == 0, 1], s = 30, c = 'salmon', label = 'Cluster 1')      #scatter plot for cluster1                
plt.scatter(X.iloc[y == 1, 0], X.iloc[y == 1, 1], s = 30, c = 'blue', label = 'Cluster 2')        #scatter plot for cluster2
plt.scatter(X.iloc[y == 2, 0], X.iloc[y == 2, 1], s = 30, c = 'green', label = 'Cluster 3')       #scatter plot for cluster3
plt.scatter(X.iloc[y == 3, 0], X.iloc[y == 3, 1], s = 30, c = 'orange', label = 'Cluster 4')      #scatter plot for cluster4     
plt.scatter(km.cluster_centers_[:, 0], km.cluster_centers_[:, 1],
            s = 200, c = 'Red', label = 'Centroid',marker = '*')                                  #scatter plot for centroid
plt.title('Clusters of customers')                                                                #Adding title to the scatter plot
plt.xlabel('Annual Income (k$)')                                                                  #Giving label for x-axis
plt.ylabel('Spending Score (1-100)')                                                              #Giving label for y-axis
plt.legend(title = 'Cluster')                                                                     #Adding legend
plt.show()


#Y = df['class']
Y = pd.DataFrame(Y)                                                        
Y['class'].replace({'van':0, 'bus':1,'saab':2,'opel':3},inplace = True)    #replacing the values with silhoutte
z = Y['class']
z

0      0
1      0
2      2
3      0
4      1
      ..
841    2
842    0
843    2
844    2
845    0
Name: class, Length: 846, dtype: int64


print(classification_report(cluster, z))                                 #printing classification report

              precision    recall  f1-score   support

           0       0.50      0.28      0.36       359
           1       0.17      0.20      0.18       188
           2       0.24      0.40      0.30       131
           3       0.17      0.21      0.18       168

    accuracy                           0.27       846
   macro avg       0.27      0.27      0.26       846
weighted avg       0.32      0.27      0.28       846


print(confusion_matrix(cluster,z))                                      #printing confusion matrix

[[100 153  57  49]
 [  0  37  73  78]
 [  0  28  53  50]
 [ 99   0  34  35]]


KMeans_Accuracy = metrics.accuracy_score(cluster, z)                   #accuracy
KMeans_Accuracy

0.26595744680851063


Accuracy = [Logistic_Regression_Accuracy, Decision_Tree_Accuracy,SVC_Accuracy, SVM_Accuracy,KMeans_Accuracy]    #storing the accuracy in a variable
Models = ['Logistic','Decision','SVM(Before Tuning)','SVM(After Tuning)','Kmeans']                              #storing the column names in variable model
compare = pd.DataFrame()                                                                                        
compare['Model'] = Models                                                                                       #previewing
compare['Accuracy'] = Accuracy
compare


check = sns.light_palette('green',as_cmap=True)                         #setting style to the cmap
display = compare.style.background_gradient(cmap=check)                 #previewing the dataframe
display


Accuracies=np.array(Accuracy)                    #initializing the numpy array and storing the accuracy in it
Accuracies=np.multiply(Accuracies,100)           #converting into percentages
Accuracies=np.round(Accuracies,2)                #rounding it 2 decimal points


plt.figure(figsize=(10,5))                                                     #resizing the graph
sns.barplot(x=Models,y=Accuracies)                                             #plotting the barplot
plt.ylim((0,110))                                                              #setting ylimit
for index,data in enumerate(Accuracies):                                       #entering into for loop for adding text to bar
    plt.text(x=index , y =data+1 , s=f"{data}%" , fontdict=dict(fontsize=10))  #adding text
plt.tight_layout()
plt.show()

	compactness	circularity	distance_circularity	radius_ratio	pr.axis_aspect_ratio	max.length_aspect_ratio	scatter_ratio	elongatedness	pr.axis_rectangularity	max.length_rectangularity	scaled_variance	scaled_variance.1	scaled_radius_of_gyration	scaled_radius_of_gyration.1	skewness_about	skewness_about.1	skewness_about.2	hollows_ratio
count	846.000000	846.000000	846.000000	846.000000	846.000000	846.000000	846.000000	846.000000	846.000000	846.000000	846.000000	846.000000	846.00000	846.000000	846.000000	846.000000	846.000000	846.000000
mean	93.678487	44.861702	82.088652	168.940898	61.693853	8.567376	168.839243	40.933806	20.582742	147.998818	188.625296	439.911348	174.70331	72.462175	6.377069	12.599291	188.932624	195.632388
std	8.234474	6.169866	15.771533	33.472183	7.888251	4.601217	33.244978	7.811560	2.592138	14.515652	31.394837	176.692614	32.54649	7.486974	4.918353	8.931240	6.163949	7.438797
min	73.000000	33.000000	40.000000	104.000000	47.000000	2.000000	112.000000	26.000000	17.000000	118.000000	130.000000	184.000000	109.00000	59.000000	0.000000	0.000000	176.000000	181.000000
25%	87.000000	40.000000	70.000000	141.000000	57.000000	7.000000	146.250000	33.000000	19.000000	137.000000	167.000000	318.250000	149.00000	67.000000	2.000000	5.000000	184.000000	190.250000
50%	93.000000	44.000000	80.000000	167.000000	61.000000	8.000000	157.000000	43.000000	20.000000	146.000000	178.500000	364.000000	173.00000	71.500000	6.000000	11.000000	188.000000	197.000000
75%	100.000000	49.000000	98.000000	195.000000	65.000000	10.000000	198.000000	46.000000	23.000000	159.000000	217.000000	587.000000	198.00000	75.000000	9.000000	19.000000	193.000000	201.000000
max	119.000000	59.000000	112.000000	333.000000	138.000000	55.000000	265.000000	61.000000	29.000000	188.000000	320.000000	1018.000000	268.00000	135.000000	22.000000	41.000000	206.000000	211.000000

	compactness	circularity	distance_circularity	radius_ratio	pr.axis_aspect_ratio	max.length_aspect_ratio	scatter_ratio	elongatedness	pr.axis_rectangularity	max.length_rectangularity	scaled_variance	scaled_variance.1	scaled_radius_of_gyration	scaled_radius_of_gyration.1	skewness_about	skewness_about.1	skewness_about.2	hollows_ratio	class
0	95	48	83	178	72	10	162	42	20	159	176	379	184	70	6	16	187	197	van
1	91	41	84	141	57	9	149	45	19	143	170	330	158	72	9	14	189	199	van
2	104	50	106	209	66	10	207	32	23	158	223	635	220	73	14	9	188	196	saab
3	93	41	82	159	63	9	144	46	19	143	160	309	127	63	6	10	199	207	van
4	85	44	70	205	103	52	149	45	19	144	241	325	188	127	9	11	180	183	bus

	Actual	Predicted
481	opel	opel
814	bus	bus
510	bus	van
219	van	van
383	opel	opel
...	...	...
675	opel	opel
226	saab	saab
449	van	van
785	bus	bus
380	opel	opel

	Model	Accuracy
0	Logistic	0.814961
1	Decision	0.795276
2	SVM(Before Tuning)	0.456693
3	SVM(After Tuning)	0.783465
4	Kmeans	0.265957

Problem statement¶

Description of Dataset¶

Vehicle Classification¶

Data Dictionary¶

Importing Required Modules¶

Read the dataset¶

Check for Missing Data¶

Exploratory Data Analysis¶

Data Preprocessing (Understanding the data)¶

Understanding the outier using different plots¶

Apply different Classification Algorithms and tune them¶

Split Data for Training and Testing¶

Logistic Regression Model¶

Training and Predicting¶

Predicting¶

Evaluation¶

Heapmap for Confusion Matrix¶

Classification Report¶

Accuracy¶

Creating DataFame to roughly compare actual outputs and predicted outputs¶

Decision Tree¶

Training and Predicting¶

Predicting¶

Model Evaluation¶

Creating DataFame to roughly compare actual outputs and predicted outputs¶

Visualising Decision Tree¶

Support Vector Classifier (SVM)¶

Predicting¶

Model Evaluation¶

Creating DataFame to roughly compare actual outputs and predicted outputs¶

Hyper Parameter Tuning using GridSearch¶

Prediction using tuned model¶

K Means¶

Elbow Plot¶

Initializing K-Means¶

Getting clusters¶

Getting clusters of final cluster¶

Scatter plot as per the clustered¶

Get performance metrics for all the applied classifiers¶

Visually compare the performance of all classifiers¶

Best possible tuning from all the classifiers¶

Future scope of improvement in the algorithm¶